An advanced approach to concrete mix proportion design: Integrating artificial intelligence with dense packing theory

The development of cost-effective concrete requires multi-objective optimization of mix proportions to balance performance, cost, and sustainability. This study presents an innovative concrete mix design methodology that integrates artificial intelligence (AI) with dense packing theory through a hybrid framework combining the CatBoost algorithm and an elite-strategy enhanced Non-dominated Sorting Genetic Algorithm II (NSGA-II). The framework demonstrates excellent predictive capability (R² = 0.977 for strength prediction). It generates optimized solutions through multi-objective optimization, with the rational point method identifying the optimal compromise among strength, cost, and carbon emissions. By incorporating dense packing theory, the method accounts for the particle size distribution of local materials, resolving gradation uncertainties while achieving 5.57 % strength improvement compared to conventional mixes, without compromising cost or efficiency in emissions. Practical validation confirms the method's effectiveness in producing superior concrete mixtures adapted to site-specific material characteristics, demonstrating significant potential for enhancing both performance and sustainability in concrete construction.